History of West Mediterranean Newts, Pleurodeles (Amphibia

Home , Chelotriton, Pleurodeles, Pleurodeles poireti

Systematics and Biodiversity 1 (3): 327–337 Issued 16 February 2004 DOI: 10.1017/S1477200003001221 Printed in the United Kingdom C The Natural History Museum

S. Carranza & E. N. Arnold∗ History of West Mediterranean newts, Department of Zoology, The Natural History Museum, Pleurodeles (Amphibia: Salamandridae), London, Cromwell Road, London, SW7 5BD, UK inferred from old and recent DNA submitted April 2003 accepted July 2003 sequences

Abstract MtDNA sequences (396 bp cytochrome b and 369 bp 12S rRNA) from recent material and old museum specimens indicate Pleurodeles poireti and P. waltl form independent clades with 7.76% genetic divergence. Within P. poireti, populations from Djebel Edough, NE Algeria are very distinct with 6.12% genetic divergence from the remainder and may deserve separate species status. Away from Djebel Edough, P. poireti consists of three distinct clades (coastal NW Tunisia; central N Algeria; Constantine plus inland NW Tunisia) with a maximum genetic divergence of only 1%. P. waltl contains two clades with 2.96% genetic divergence, one in SE and E Spain plus north Morocco, the other in Portugal and SW and central Spain. Pleur- odeles probably invaded NW Africa from SW Europe during the Messinian Salinity Crisis, when land contact was first established at 5.6 Ma, and then interrupted at 5.3 Ma. Molecular clocks, calibrated in the assumption that the latter event separated P. waltl and P. poireti, suggest that Pleurodeles diverged from its sister taxon, Tylototriton, at about 8.6–10 Ma. Djebel Edough P. poireti separated at about 4.2 Ma, perhaps through isolation on a temporary, now ‘fossil’, island initiated by the Messinian crisis. Differentiation in remaining P. poireti may have been caused by Pleistocene climatic fluctuations, while bifurcation in P. waltl appears to have taken place in the Pliocene approximately between 3.2 and 2 Ma. This species reached Morocco very recently, perhaps as a result of human introduction. Use in Pleurode- les of the slower divergence rates estimated in some other salamandrids results in a less parsimonious historical hypothesis that does not fit known geophysical events.

Key words ancient mtDNA, Pleurodeles, Tylototriton, Chelotriton,dispersal,land bridge, Messinian, fossil island, Djebel Edough

Introduction often traces of this structure on ribs 1 and 2), and sharp rib tips that project through a row of glandular swellings on the In this paper we review the available fossil and other evidence ﬂanks. P. waltl is common and widely distributed in the south- for the relationships and minimum age of the West Mediter- ern two-thirds of the Iberian Peninsula and occurs with less ranean salamandrid newt genus Pleurodeles, and use mtDNA abundance in northern Morocco, where it is found in the area sequence to investigate the detailed phylogeny of the group. delineated by Tangier, Casablanca and Alhoceima (Bons & The two sources of information are then employed to speculate Geniez, 1996). The second Pleurodeles species, P. poireti,is about the history of Pleurodeles including the origin of its small (up to about 200 mm in total length; Pasteur, 1958), with present biogeographical pattern. 14 presacral vertebrae, no obvious tubercular processes on the ribs which are not sharp, and no glandular swellings on the Current taxonomy of Pleurodeles ﬂanks. It occurs in northern Tunisia and Algeria extending This genus has two currently recognized species. P. waltl is westwards as far as Oran. large (up to 300 mm in total length; Pasteur, 1958), with usually 15 presacral vertebrae, a tubercular process on rib 3 (and Relationships and possible age of Pleurodeles Morphology and mitochondrial DNA sequences (a total of about 1000 bp of 12S rRNA, 16S rRNA and valine transfer *Corresponding author. Email: [email protected] RNA) indicate that among living forms, the sister taxon of

327 328 S. Carranza & E. N. Arnold

Pleurodeles is the species of Tylototriton in South-East Asia, waltl from the Iberian Peninsula and Morocco and 23 P.poireti including the type species of that genus, T. verrucosus (Titus & from Algeria and Tunisia, while individuals of Salamandra sa- Larson, 1995). Within the Salamandridae, Pleurodeles and lamandra morenica, S. s. terrestris, Tylototriton taliangensis Tylototriton form an assemblage with a number of European and Mertensiella luschani were used as outgroups. Sequences fossil taxa (Group II of Estes, 1981), which is characterized of the last two species were obtained from Genbank. Ten of the by extensive sculptured dermal bone on the skull and some- P. poireti investigated were acquired between 1878 and 1888 times on the neural spines of the presacral vertebrae, complete by Ferdinand Lataste who bequeathed them, with the rest of frontosquamosal arches on the skull, and long ribs, at least his large herpetological collection, to George A. Boulenger at some of which typically have dorsal tubercular processes. One the Natural History Museum, London, where they were acces- genus in particular, the extinct Chelotriton, is very similar to sioned in 1920. Lataste apparently preserved his specimens in Tylototriton and was widespread in Europe from the Eocene a distilled wine product, probably brandy. To prevent evapor- to the Pliocene (Estes, 1981; Bailon, 1989). Both these taxa ation, he stored them in containers closed with a glass disc have extensive dermal bone on the skull and forming expanded attached with putty; this was then covered with parchment plaques on the neural spines of the vertebrae. Living Tyloto- that was tied with twine and finally sealed with black paint. In triton differ from Chelotriton in their smaller body size, an some cases, Lataste’s containers have been left unopened to the ornamentation of pits and grooves on the cranial osteoderms present day, so the enclosed specimens have remained in what instead of pustules (though some pitting may also occur in is essentially ethanol instead of being transferred to industrial Chelotriton robustus; Estes, 1981), typically 15–16 presacral methylated spirits (IMS), the usual preservative at the Natural vertebrae instead of 13–14, a tubercular process present only History Museum. Because of this there was the possibility on rib 3, ossified carpus and tarsus in at least large animals, that Lataste’s sealed specimens might still be viable sources of and a narrower head. DNA, which is usually largely or wholly degraded by IMS. To Tylototriton has been reported from the middle Eocene avoid contamination, Lataste’s animals were treated as ancient of Germany, as T. weigelti Herre 1935. Individuals assigned to DNA samples. The opening of the bottles, handling of speci- this taxon appear to have dermal ossification similar to that of mens, DNA extraction and PCR amplification were all carried modern Tylototriton, but the type is very small and all available out in isolation, following procedures previously successfully specimens are incomplete. This means several critical features used in other attempts to obtain DNA from old Museum speci- cannot be checked, including presacral number and degree mens (Carranza et al., 1999, 2001; Carranza & Arnold, 2003). of ossification of the carpus and tarsus. T. weigelti may in Nine of the ten Lataste specimens investigated yielded ad- fact be a Chelotriton (Milner, 2000). The earliest specimens equate DNA that was not seriously degraded, even though assigned to Pleurodeles are quite recent, coming from the they were collected well over a century ago. DNA was ex- Upper Miocene or Lower Pliocene of Spain (Sanchiz, 1977). tracted from both Lataste and recent samples using the Tissue It has been suggested (Herre, 1941) that the genus is related to Extraction Kit from Quiagene. Primers used in both ampli- Palaeopleurodeles Herre 1941 from the Upper Oligocene of fication and sequencing were: S1F (5- TTC AAC TAC AAA Germany, but several critical features cannot be seen in known AACCTAATGACCC-3) and cytochrome b2 (Kocher specimens of Palaeopleurodeles, so this association cannot be et al., 1989) for cytochrome b, and 12Sa and 12Sb (Kocher substantiated (see Estes, 1981; Milner, 2000). et al., 1989) for 12S rRNA. The molecular evidence for the sister relationship of Pleurodeles and Tylototriton is also supported by shared mor- Phylogenetic analyses phological features that are apparently derived within the Two data sets were used in the phylogenetic analyses. Data set Group II salamandrids, including an ossified carpus and tarsus, I involved all 42 specimens of Pleurodeles from Table 1, plus and sometimes 15 presacral vertebrae. The sister relationship Mertensiella luschani from Genbank and included 396 bp of indicates that the exclusive Pleurodeles lineage too may pos- the cytb and 369 bp of the 12S rRNA genes. Data set II involved sibly have arisen in the Upper Miocene. Its autapomorph- a representative of each of the six main Pleurodeles clades that ies include reduction of osteodermal bone (completely so on were discovered by analysis of data set I (highlighted in Figs 1 the vertebrae) and in the size of the tubercular processes and 2), plus S. s. terrestris, S. s. morenica, M. luschani and on the ribs. Comparison with Tylototriton and other sala- Tylototriton taliangensis; it included 346 bp of the cytb gene. mandrids indicates that the very large size and sharp ribs Analysis of this second data set was necessary to permit the of P. waltl are derived, as are the following features of P. inclusion of Tylototriton, the sister of Pleurodeles, because the poireti: small size, 14 presacral vertebrae, and lack of tuber- areas of sequence available for this taxon are different from cular processes and of glandular swellings on the side of the the other samples. Analysis of data set II provided a check that body. tree topology of Pleurodeles was not changed by inclusion of its closest relative and permitted an estimate to be made of the age of the node at which Pleurodeles and Tylototriton Material and Methods separated. DNA sequences were aligned using ClustalX (Thompson Samples and DNA extraction et al., 1997) with default parameters. No gaps were included A total of 46 specimens of the family Salamandridae were in the cytochrome b (cytb) alignment and no stop codons were used in this study (see Table 1 and Fig. 1). They included 19 P. observed when the sequences were translated into amino acids Phylogeography of Pleurodeles 329 Museum, London; Lat. Coll – Lataste AY222532/AY222488 E181225 AY222543/AY222499 E18129 AY222523/AY222479AY222524/AY222480 E181216 E181217 AY222542/AY222498 E18128 AY222530/AY222486 E181223 AY222535/AY222491AY222538/AY222494 E18123 E18124 AY222541/AY222497AY222526/AY222482 E18127 AY222517/AY222473AY222539/AY222495 E18122 E18121 E18125 AY222533/AY222489 E181228 AY222527/AY222483AY222528/AY222484 E181220 E181221 AY222521/AY222477 E181213 AY222540/AY222496 E18126 AY222525/AY222481 E181219 AY222522/AY222478 E181214 AY222534/AY222490 E181229 AY222529/AY222485 E181222 AY222536/AY222492 E181230 AY222514/AY222470 E10113 AY222518/AY222474AY222519/AY222475AY222520/AY222476 E181210 E181211 E181212 AY222531/AY222487 E181224 AY222503/AY222459AY222516/AY222472 E171249 E10115 AY222537/AY222493AY222512/AY222468AY222513/AY222469 E181231 E10111 E10112 AY222502/AY222458 E171243 AY222515/AY222471 E10114 Genbank Accession Genbank uller, 1881 AY222507/AY222463 Anc13 uller, 1881 AY222508/AY222464 Anc14 uller, 1882 AY222505/AY222461 Anc11 adiz (Spain) adiz (Spain) adiz (Spain) ordoba (Spain) en (Spain) adiz (Spain) azaro, Villamanrique de la Condesa; Sevilla en (Spain) ın de Valdeiglesias; Avila (Spain) o; Tarragona (Spain) ıa (Algeria) Lat. Coll. 1920.1.20.1750 Hagenm¨ (Spain) elmez/Hinojosa; C´ La Parra (Badajoz) Badalona (Barcelona) Details of material and sequences used in the present study. Numbers after the species names refer to localities shown in Fig. 1. BMNH – Natural History Collection. –20 Annaba, Djebel Edough (Algeria) Lat. Coll. 1920.1.20.1327 Hagenm¨ –21 Annaba, Djebel Edough (Algeria) Lat. Coll. 1920.1.20.1369 Hagenm¨ –26 5 Km South of Tabarca (Tunisia) 36.90778S/8.74952W –22–23–24–25 Tabarca (Tunisia) Tabarca (Tunisia) Tabarca (Tunisia) 5 Km. West of Nefza (Tunisia) –27 5 Km South of Tabarca (Tunisia) 36.90778S/8.74952W –28–29–30–31–32 Larba (Algeria)–33 Beja¨ –34 Larba (Algeria)–35 Alger (Algeria) Theniet El-Had (Algeria) Theniet El-Had (Algeria) Constantine (Algeria) Jendouba (Tunisia) 36.4941S/8.76945E Lat. Coll. 1920.1.20.3824 Lat. Coll. 1920.1.20.1529 Lat. Coll. Lat. 1920.1.20.1511 Coll. 1920.1.20.2539 Lat. Coll. 1920.1.20.18 Lat. Coll. 1920.1.20.1504 Lallemant, 1888 Bedel, 1884 Bedel, Lallemant, 1884 1881 Henon, 1881 Lasserre, 1878 AY222504/AY222460 AY222510/AY222466 AY222544/AY222500 AY222545/AY222501 Anc10 AY222511/AY222467 AY222462/AY222506 Anc8 E30091 E30092 Anc9 Anc12 –36–37–38–39 Near Fernana (Tunisia) 36.64087S/8.69703E Dam Bou Heurtma (Tunisia) 36.6739S/8.75785E Dam Bou Heurtma (Tunisia) 36.6739S/8.75785E Ain Draham (Tunisia) 36.70915S/8.68623E –40–41–42 Dam Bou Heurtma (Tunisia) 36.6739S/8.75785E Dam Bou Heurtma (Tunisia) 36.6739S/8.75785E Dam Bou Heurtma (Tunisia) 36.6739S/8.75785E –18 Puerto Alazores; Granada (Spain) –19 Tanger (Morocco) –17 Lag. de San L´ –16 South of Laguna de Medina; C´ –11–12–13–14–15 Charf-la-Kaab, Tanger (Morocco) 37.77155S/5.78785W Near Jerez de La Frontera; C´ Quesada/Jodar, Ja´ Near Jerez de La Frontera; C´ Torregorda; C´ –7–8–9–10 Marshes of Charf-la-Kaab, 10 miles from Morhane Tanger (Morocco) (Morocco) Morhane (Morocco) Charf-la-Kaab, Tanger (Morocco) 37.77155S/5.78785W Lat. Coll. 1920.1.20.3591 Vaucher, 1887 AY222509/AY222465 Anc18 –1–2–3–6 Extremos (Portugal) Aroche; Huelva (Spain) San Mart´ Genave; Ja´ –4–5 B´ Perell´ P. poireti P. waltl P. waltl P. poireti P. waltl P. poireti P. poireti P. poireti P. poireti P. poireti P. waltl P. poireti Table 1 P. waltl P. waltl P. waltl P. waltl P. waltl P. waltl P. waltl P. waltl P. waltl P. poireti P. poireti P. poireti P. poireti P. poireti P. poireti P. poireti P. poireti TaxaS. s. morenica S. s. terrestris Locality BMNH number Collector and Year number (Cytb/12S) sequence code P. waltl P. waltl P. waltl P. waltl P. poireti P. poireti P. poireti P. poireti P. waltl P. waltl P. poireti P. poireti P. poireti 330 S. Carranza & E. N. Arnold

Figure 1 Map of the Iberian Peninsula and North Africa showing localities of Pleurodeles samples used in the present study, see Table 1 and Fig. 2 for further details. The Guadalquivir River has been highlighted in black. using the vertebrate mitochondrial code. This suggests that all as part of the analysis and four incrementally heated Markov the cytb sequences analysed were functional. All the Pleur- chains with the default heating values. All analyses started odeles 12S rRNA sequences were of the same length and only with randomly generated trees and ran for 2.5 × 106 genera- two gaps had to be postulated to align the three outgroup se- tions, with sampling at intervals of 100 generations that pro- quences with the ingroup. duced 25 000 sampled trees. To ensure that the analyses were Three methods of phylogenetic analysis were em- not trapped on local optima, the data set was run three times ployed for both data sets and their results compared. These independently, each run beginning with a different starting were maximum-likelihood (ML), Bayesian analysis, and tree. The log-likelihood values of the 25 000 trees in each ana- Maximum-parsimony (MP). Modeltest (Posada & Crandall, lysis were plotted against the generation time. All the trees 1998) was used to select the most appropriate model of se- produced prior to reaching stationarity were discarded, mak- quence evolution for the ML and Bayesian analyses, under the ing sure that burn-in samples were not retained. Although Akaike Information Criterion. For both data sets this was the stationarity was reached very rapidly (data not shown), only General Time Reversible model (GTR) taking into account the last 5000 trees in each of the three independent analyses the shape of the Gamma distribution (G). Bayesian phylo- were used to estimate separate 50% majority rule consensus genetic analyses were performed with MRBAYES v. 2.01 trees for these. The frequency of any particular clade, among (Huelsenbeck & Ronquist, 2001) using the GTR+G model the individual trees contributing to the consensus tree, repres- of sequence evolution (see above) with parameters estimated ents the posterior probability of that clade (Huelsenbeck & Phylogeography of Pleurodeles 331

Figure 2 ML tree for Pleurodeles (Log likelihood − 2291.95312, GTR+G model of sequence evolution) inferred from data set I. Bootstrap support and posterior probabilities for particular nodes are shown in the boxes, with the ﬁgures indicating the percentage support for different analyses. Upper left, bootstrap support derived by ML (GTR+G). Upper middle, bootstrap support derived by MP (ts = tv). Upper right, bootstrap support derived by MP (ts = 1; tv = 6). Lower left, posterior probability values derived from Bayesian analysis (1st replicate). Lower middle, posterior probability values derived from Bayesian analysis (2nd replicate). Lower right, posterior probability values derived from Bayesian analysis (3rd replicate). Estimated dates are given for most important bifurcations, which are marked by ﬁlled circles. Dates above the line are derived from Kimura 2-parameter genetic distances, those below are calculated using the NPRS method implemented in r8s (see Material and Methods).

Ronquist, 2001); only values above 95% were regarded as ratio of data sets I and II. Nodal support of the ML and MP indicating that clades were signiﬁcantly supported. results was assessed by bootstrap analysis (Felsenstein, 1985) Both ML and MP analyses were performed in PAUP* involving 1000 pseudo-replications. The incongruence length 4.0b10 (Swofford, 1998) and included heuristic searches in- difference (ILD) test (Mickevich and Farris, 1981; Farris et al., volving tree bisection and reconnection (TBR) branch swap- 1994) was used to check for incongruence between the genes ping with 100 random stepwise additions of taxa. Gaps were used. In this test, 10 000 heuristic searches were made and in- included as a ﬁfth state. In the MP analyses of data set I, variable characters were removed before starting the analysis transversions were given the same weight as transitions and (Cunningham, 1997). six times that weight. For data set II transversions were given Because it was intended to establish approximate dates the same weight and nine times that weight. This weighing for events on the phylogeny of Pleurodeles, the likelihood scheme was applied in order to correct for the observed ts/tv ratio test (Huelsenbeck & Crandall, 1997) was used to assess 332 S. Carranza & E. N. Arnold

Figure 3 ML tree for Pleurodeles (Log likelihood –1244.90457, GTR+G model of sequence evolution) inferred from data set II. Bootstrap support and posterior probabilities for particular nodes are shown in the boxes with the ﬁgures indicating the percentage support for different analyses. Upper left, bootstrap support derived by ML (GTR+G). Upper middle, bootstrap support derived by MP (ts = tv). Upper right, bootstrap support derived by MP (ts = 1; tv = 9). Lower left, posterior probability values derived from Bayesian analysis (1st replicate). Lower middle, posterior probability values derived from Bayesian analysis (2nd replicate). Lower right, posterior probability values derived from Bayesian analysis (3rd replicate). Estimated dates are given for some bifurcations, which are marked by ﬁlled circles. Dates above the line are derived from Kimura 2-parameter genetic distances, those below are calculated using the NPRS method implemented in r8s (see Material and Methods).

the statistical signiﬁcance of the difference between the log for the MP trees when uninformative sites were excluded are: likelihood of the trees calculated with and without molecular MP (tv = 6ts= 1): two trees of 488 steps, CI (consistency in- clock assumptions. Absence of signiﬁcant difference would index) = 0.867, RI (retention index) = 0.975; MP (tv = ts): eight dicate gene evolution within the phylogeny was clocklike. Di- trees of 218 steps, CI = 0.794, RI = 0.964. vergence times on trees were also estimated using the Nonpara- The most basal bifurcation separates the populations metric Rate Smoothing (NPRS) method (Sanderson, 1997) presently placed in P. poireti from those assigned to P. waltl implemented in the program r8s. The topology used to calcu- (7.76% genetic divergence for cytb + 12S rRNA genes). Within late the dates here were the ML trees shown in Figs 2 and 3 the P. poireti clade there is a deep division (6.12% genetic di- with GTR+G branch lengths. vergence) between samples from Annaba (previously Bone)ˆ in the Djebel Edough region, on the northeast Algerian coast 70 km from the Tunisian border (clade 3, Fig. 2), and all Results other material (clades 4–6). The degree of genetic divergence found between Djebel Edough Pleurodeles and all the rest of Analysis of data set I P. poireti analysed approaches that separating the two spe- A total 765 bp (396 bp of cytb and 369 bp of 12S rRNA) cies currently recognized in the genus. It may consequently be were used in the analyses. Of these, 181 bp were variable and appropriate to assign species status to the Djebel Edough pop- 146 parsimony-informative. The ILD test showed that the two ulation as well, especially as it appears to be morphologically gene fragments were congruent (ILD P > 0.53) and could con- distinct (S. Carranza & E. O. Z. Wade, unpubl. data). sequently be combined in total evidence analyses. The results The remaining P. poireti form a well-supported mono- of these are summarized in Fig. 2. ML and Bayesian ana- phyletic group made up of three distinct units with a maximum lyses produced almost identical topologies that recognized all genetic divergence of only 1% (Table 2). Clade 4 contains six major clades highlighted in Fig. 2. The only differences populations to the north of the Medjerda Mountains in coastal were slight variations in the arrangement of samples within northwest Tunisia; clade 5 ones from central north Algeria, and these clades. The two MP analyses, involving different relative its sister clade 6 those from Constantine in northeast Algeria weightings of transversions (tv) and transitions (ts) produced (locality 34, Fig. 1) and from inland northwest Tunisia (in and trees that recognized all six main clades shown in Fig. 2. Values south of the Medjerda Mountains). Phylogeography of Pleurodeles 333

Geographical distance between K2P mean genetic distances studied localities (Km)

Cytb+12S Cytb 12S maximum minimum Within Clade 1 0.00 0.00 0.00 350 100 Within Clade 2 0.20 0.19 0.21 950 20 Within Clade 3 0.00 0.00 0.00 0 0 Within Clade 4 0.04 0.00 0.09 25 5 Within Clade 5 0.06 0.00 0.12 300 20 Within Clade 6 0.08 0.05 0.12 195 3 Within Morocco 0.31 0.35 0.27 150 4 P. waltl Clades 1–2 vs P. poireti Clades 3–6 7.76 10.38 5.06 1500 600 P. waltl Clade 1 vs P. waltl Clade 2 2.96 4.60 1.27 727 95 P. poireti Clade 3 vs P. poireti Clade 4–6 6.12 8.35 3.82 500 85 P. poireti Clade 4 vs P. poireti Clade 5–6 1.00 0.75 1.25 636 18 P. poireti Clade 5 vs P. poireti Clade 6 0.82 0.73 0.91 620 142

Table 2 Kimura 2-parameter genetic distances (%) within and between populations (see Fig. 1).

Within P. waltl there is also a bifurcation (2.96% genetic lyses are shown in Fig. 3. Maximum Parsimony, Maximum divergence). This separates animals from the southeast and Likelihood and Bayesian analysis all produced identical trees. east of Spain plus northern Morocco (clade 2) from those in Among the restricted number of salamandrid taxa included, Portugal and southwest and central Spain (clade 1). All the Tylototriton is sister to Pleurodeles with high bootstrap sup- six distinct clades within Pleurodeles are genetically quite port and posterior probability values, in agreement with ana- homogeneous across large geographical distances (see Figs 1, lyses involving all salamandrid data (Titus & Larson, 1995). 2 and Table 2). Despite the restricted amount of sequence used, the topology When the log likelihood value of the ML (GTR+G) within Pleurodeles is identical to that obtained from analysis tree from Fig. 2 (−2291.95312) was compared with the of data set I (Fig. 2). Values for the MP trees excluding un- log likelihood of the same tree constructed under molecu- informative sites are as follows: MP(ts = tv) two trees of 151 lar clock assumptions (−2300.7860), there was no signific- steps, CI = 0.669, RI = 0.718; MP(tv = 9, ts = 1) one tree of ant difference between the two (likelihood ratio test statistic, 423 steps, CI = 0.7, RI = 0.845. The log likelihood of the ML − = 2 2log 17.6658, which approximates to a χ43 distribution tree from Fig. 3 was not significantly different to the log like- under the null hypothesis; P < 0.05). The ML tree can there- lihood of the ML tree inferred under the assumption of a mo- − = 2 fore be used for estimating dates. The Kimura 2-parameter lecular clock ( 2log 4.07, which approximates to a χ8 (K2P) correction was used, to make results comparable with distribution under the null hypothesis; P < 0.05) and therefore other genetic distances and clock calibrations of other groups the short cytb fragment could be used to estimate the approx- previously used by the authors (Carranza et al., 2000; Rando imate age of the split between Tylototriton and Pleurodeles. et al., in press). Relevant K2P genetic distances between the The calibration point employed was the same as for data set six main clades, and the genetic variability within them are I (see above, and Fig. 2). The rate of molecular evolution shown in Table 2. for the 346 bp of cytb analysed was 2.0% per million years, The separation of African P. poireti from the essentially slightly higher than the rate calculated for the same gene using Iberian P. waltl is used as a calibration point and, for reas- data set I. This difference probably results from the different ons given later, this event is dated at 5.3 M.y. On the basis amounts of cytb sequence used in data sets I (396 bp) and II of the 7.76% difference between the combined cytb and 12S (346 bp). rRNA fragments of the two species, the mean divergence rate is estimated as 1.46% per million years. When the two gene fragments are treated independently, the rates are respect- Discussion ively 1.95% and 0.95%. The Nonparametric Rate Smoothing (NPRS) method produced a similar date to that based on the The relationship of Pleurodeles successively to Tylototriton Kimura 2-parameter genetic distances (Fig. 2) for the split and to other members of the Group II assemblage of salaman- between the Djebel Edough and the other P. poireti popul- drids, which has been present in Europe as far back as the ations, but dates for later events were distinctly older. Eocene, indicates that it arose in Europe and only later spread into its African range. As fossils suggest Pleurodeles may have arisen only late in the Miocene, there are two primary routes Analysis of data set II by which it may have spread into Africa. One is via the land Out of the 346 bp of cytb aligned 111 were variable and 80 bridge between Europe and Africa that has existed more or parsimony-informative. The results of the phylogenetic ana- less continuously from the mid-Miocene. This arose when the 334 S. Carranza & E. N. Arnold eastern end of the Mediterranean closed at 19–15 Ma, as a The distinct nature of Djebel Edough Pleurodeles could result of the northeast part of the African plate colliding with have arisen in a number of ways, but the simplest would be by the Turkish plate (Adams et al., 1983; Rogl,¨ 1999), permit- the massif being isolated from the rest of Algeria long enough ting biotic interchange between nearby areas of Eurasia and for speciation to occur. This is possible as the present geo- Africa. graphy of Djebel Edough indicates it may be a fossil island. The second route involved a transient land contact at the Fossil islands consist of areas of land that were once isolated other extremity of the Mediterranean Sea between the Iberian by sea but later became continuous with other usually larger Peninsula and northwest Africa, right at the end of the Mio- land masses. In the Mediterranean area, they originated at vari- cene. This was responsible for the event known as the Messin- ous times and were often inhabited by endemic animals. For ian Salinity Crisis. About 5.59 Ma, tectonic uplift of more example, Monte Gargano, which is now part of the Adriatic than 1000 m along the African and Iberian continental margins coast of Italy, was isolated for a considerable period beginning formed the Gibraltar arch producing a land bridge. This closed in the Miocene, during which time a distinctive fauna evolved, the two marine gateways between the Atlantic Ocean and the including murine rodents, Microtia (Giuli & Torre, 1984) and Mediterranean Sea that existed in the Miocene (Duggen et al., a giant hedgehog, Deinogalerix (Butler, 1980). Other islands 2003) and isolated the Mediterranean. Without input from the were formed in the Messinian. Rivers flowing into the Medi- Atlantic, its surface level fell by over 1000 m, perhaps in terranean in this period cut deep canyons (Said, 1990, 1993) less than 1000 years (Hsu,¨ 1972; Hsu¨ et al., 1973; Blondel & which occasionally separated a coastal area from its hinter- Aronson, 1999; Krijgsman et al., 1999a). The fall desiccated land (Hsu,¨ 1972; Hsu¨ et al., 1973, 1977; Blondel & Aronson, large areas of the Mediterranean Sea bed which were sub- 1999), for instance if the canyons of adjacent rivers coalesced. sequently partly covered with freshwater sediments brought in When the Mediterranean basin filled again at the end of the by rivers. The end of the Messinian Salinity Crisis at 5.3 Ma Messinian, such coastal areas sometimes became real islands was caused by the collapse of the Gibraltar arch which opened as the encircling canyons filled with salt water. On occasion, the Strait of Gibraltar. This allowed the entire Mediterranean the canyons subsequently became choked with fluvial deposits, basin to fill again in less than 100 yr (Hsu,¨ 1972; Hsu¨ et al., rejoining the islands to their hinterlands and producing a char- 1973, 1977; Blondel & Aronson, 1999; Krijgsman et al., acteristic topology in which a coastal massif is separated from 1999a) and broke the land connection which had existed for high ground further inland by rivers and their low-lying plains. nearly 300 k.y. This is true for the Djebel Edough massif which is divided from Of the two possible routes into northwest Africa, the first the rest of Algeria by lowlands that contain streams, marshes is unlikely to have been used by Pleurodeles. There is no and the 13 km by 10 km Lake Fzara. However, the molecular evidence, either in the form of fossils or relict populations, clock derived from data set I indicates that the actual separation that this genus ever existed east of the Iberian Peninsula or of the Djebel Edough population occurred at 4.2 Ma, some time northwest Africa. In other amphibian groups that occur more after the end of the Messinian Salinity Crisis. So perhaps any widely in the Mediterranean area and have been adequately canyons around Djebel Edough were not initially entirely con- studied, the closest relatives of northwest African populations tinuous and final land connection was only broken rather later. are in the Iberian Peninsula, making it the most likely source All the other P. poireti populations (clades 4–6) are ge- area. This is true, for example for Discoglossus (Garcia-Par´ıs & netically similar despite having an ancestral lineage that is an Jockusch, 1999), Alytes (Arntzen & Garcia-Par´ıs, 1995), Rana estimated 4.2 Ma old. Their high genetic similarity may res- (Beerli et al., 1996), Pelobates and Bufo bufo (S. Carranza & ult from this lineage occupying a very restricted geographical E. N. Arnold, unpubl. data). area for most of its history, only spreading and diversifying For these reasons, it is hypothesized that Pleurodeles over north Algeria and Tunisia quite recently. Alternatively, entered northwest Africa across the Gibraltar land bridge and the ancestral lineage may have been widespread since its ori- gene flow between P. waltl and P. poireti was interrupted when gin, occupying many of the humid parts of the Maghreb where the Mediterranean Sea refilled at 5.3 Ma. This event is used to the only other urodele is Salamandra algira. If so, it must calibrate the molecular clocks employed in the present study have formed an essentially panmictic unit in which genetic (p. 333). The clock based on data set II indicates that Pleur- homogeneity was maintained by gene flow between at least odeles and Tylototriton separated at 8–10 Ma, a date compat- intermittently continuous populations. On present evidence, it ible with available fossil evidence, which suggests an origin is not possible to decide between these alternatives, although for the clade containing these genera in the Late Miocene. the second seems more plausible. As parts of the desiccated Mediterranean sea floor Whatever its previous history, this lineage underwent were covered with freshwater sediments, the ancestor of the divergence in the Pleistocene, separating into three units P.poireti clade may have been able to cross directly from south- (clades 4, 5 and 6). As noted, the most divergent populations east Spain to western Algeria, rather than via the land bridge (clade 4) occur north of the Medjerda mountains (Pasmans into Morocco. Once separated from P. waltl,theP. poireti et al., 2002). They differ genetically by more than 1% from clade developed its characteristic morphological autapo- populations in and south of these mountains (clade 6), the morphies (see p. 328) before bifurcating into a very widely nearest of which is less than 18 km away. As well as being ge- distributed monophyletic unit made up of the great majority of netically distinct, the populations of clade 4 are morphologic- P. poireti populations (clades 4–6; see Figs 1 and 2), and that ally differentiated, being significantly smaller than P. poireti on Djebel Edough (clade 3). from further south in Tunisia (Pasmans et al., 2002). Clade Phylogeography of Pleurodeles 335

6 also extends at least about 200 km westwards into eastern the monophyly of Euproctus has not been firmly established Algeria,toConstantine.Itismostcloselyrelated(0.82%genetic and other taxa (Triturus spp. and Neurergus) may actually be divergence) to clade 5, which occurs in central north Algeria included in the clade (Caccone et al., 1994, 1997; Steinfartz and has known populations 140–620 km west of Constantine. et al., 2002), raising the possibility that later, unappreciated It is likely that the Medjerda Mountains acted as a bar- nodes may really represent the vicariance event. In the second rier for some time in the Pleistocene (between 0.7–1.5 Ma). case, the fossils used for calibration of Notophthalmus are They probably interrupted gene flow between clade 4 and the isolated vertebrae (Estes, 1981), so identity is not necessar- rest of the populations in Tunisia and Algeria that constitute ily strongly corroborated. There is also the potential problem clades 5 and 6, which themselves separated from each other that such ancient events might not be appropriate for calibrat- at a rather later date in the Pleistocene. This separation may ing relatively fast evolving mitochondrial genes that may have have been caused by the development of a geographical barrier become saturated over the periods concerned (Avise, 2000). or may have resulted from one or more of the climatic fluctu- Finally, Euproctus and Notophthalmus–Taricha are not very ations frequent in the western Palaearctic during the Quatern- closely related to Pleurodeles (Titus & Larson, 1995). This ary. These promoted genetic and morphological differentiation makes calibration based on events within the Pleurodeles clade by changing the distribution and demography of many species preferable, as rates may vary among distant taxa. (Hewitt, 1996, 2000; Veith et al., 2003). As the maximum If, in spite of these reservations, the slower rates for the geographic distance between clades 5 and 6 is only 120 km cytb gene are applied to the Pleurodeles phylogeny, the initial (Table 2), it is unlikely that their reciprocal monophyly is due split between the P. waltl and P. poireti clades would have to a sampling artefact. Any populations subsequently found in occurred at roughly 14 Ma or more. At this time, there was the intervening area may elucidate the causes of the separation no direct land connection between the Iberian Peninsula and of these clades. northwest Africa. Southern Spain and Morocco were quite Like the P. poireti clade, P. waltl also developed its own close to each other, with a set of islands that would later become morphological autapomorphies (see p. 328) and then under- the Betic and Rifian mountains between them. But the islands went bifurcation, although this was at a later date, between 3.2 were separated from the mainlands by two main sea corridors. and 2 Ma. The split produced one form in western and central The Betic corridor in the north divided the Iberian mainland Iberia (clade 1), and another in the southeast and east of Spain from the “Betic island”, and the Rifian corridor in the south di- (clade 2), which also occurs in northern Morocco. Bifurca- vided the “Rifian island” from northern Morocco (Weijermars, tions also appear to have taken place within the same area 1988, 1991; Krijgsman et al., 1999a, 1999b, 2000; Krijgsman of the Iberian Peninsula in several other groups of amphibi- & Langereis, 2000; Duggen et al., 2003). The Betic corridor ans and reptiles including Salamandra (Garcia-Par´ıs et al., only closed at about 7.2 Ma (Krijgsman et al., 1999a, 2000), 1998), Discoglossus (Garcia-Par´ıs & Jockusch, 1999), Alytes while the Rifian corridor was only obliterated more than 1 m.y. (Arntzen & Garcia-Par´ıs, 1995) and Lacerta lepida and Psam- later (Krijgsman et al., 1999a; Krijgsman & Langereis, 2000). modromus (Carranza & Arnold, unpubl. data). There is consequently no evidence of events at 14 Ma The near identity of Moroccan P. waltl populations with that would have enabled Pleurodeles to extend southwards those from southeastern and east Iberia indicates that their an- into an area outside the Iberian peninsula and become isolated cestors reached Africa comparatively recently and long after there as the ancestral population of P. poireti. If this lineage the disappearance of the land connection associated with the had reached Morocco, there would have been a population Messinian Salinity Crisis. Because amphibians are very un- of P. poireti that became extinct in this area after spreading likely to cross salt water, the Moroccan populations may result eastwards into Algeria. An alternative possibility is that P. from anthropogenic introductions that could have been ac- waltl and P. poireti separated within the Iberian Peninsula but cidental. The Lataste specimen collected in the marshes of there is no evidence for this and it would again imply extinction Charf-la-Kaab in 1887 by Vaucher (see table 1) and identified of the original P. poireti population. as part of clade 2 in our study, confirms that individuals of Applying the slower rate of divergence, the division of P. waltl genetically very similar to the populations from the the Djebel Edough population from the rest of P. poireti would southeastern and eastern Iberian Peninsula have been present have taken place at approximately 12 Ma, at a time when there in Morocco for well over a century. was no obvious geophysical mechanism that might have pro- The calibration used above gives rates of genetic di- duced this. On the same basis, the separation of the P. waltl vergence considerably faster than those estimated for other clade into two lineages would have occurred at roughly 6 Ma salamandrids. In Euproctus newts,thesameregionofcytb around the time of the Messinian Salinity Crisis. It could be employed here was calibrated against a vicariance event, the hypothesized that the division was caused by P. waltl entering Oligocene separation of Corsica and Sardinia from southern Morocco at this period and then being isolated to produce a France, which is assumed to have isolated the ancestor of second form that then reached Spain very recently. If so, it the endemic species on these islands from that of their main- must have spread extremely rapidly and extensively there, per- land congener. This gave a divergence rate of 0.7% per Myr haps replacing its sister form, which remained in Iberia, over (Caccone et al., 1997). A similar, maximum figure for the cytb large areas. fragment was also obtained for the separation of the North In summary, using a slower rate of gene evolution than American Taricha and Notophthalmus, using fossils of known that calibrated with the Messinian Salinity Crisis, results in a stratigraphy for calibration (Tan & Wake, 1995). However, historical hypothesis that is much less parsimonious than the 336 S. Carranza & E. N. Arnold

one outlined earlier in this discussion. It involves additional CARRANZA,S.,ARNOLD, E.N., MATEO,J.A.&LOPEZ´ -JURADO, events including an extinction, two colonizations of North L.F. 2000. Long-distance colonization and radiation in gekkonid Africa instead of one, and possible partial species replacement. lizards, Tarentola (Reptilia: Gekkonidae), revealed by mitochondrial DNA sequences. Proceedings of The Royal Society of London Furthermore, most bifurcations in the phylogeny do not cor- Series B Biological Sciences 267, 637–649. relate with known geophysical events that could have caused CARRANZA,S.,ARNOLD, E.N., MATEO,J.A.&LOPEZ´ -JURADO, them. L.F. 2001. Parallel gigantism and complex colonization patterns in Cape Verde scincid lizards Mabuya and Macroscincus (Reptilia: Scincidae) revealed by mitochondrial DNA sequences. Proceed- Acknowledgements ings of The Royal Society of London Series B Biological Sciences 268, 1595–1603. We are indebted to J. Roca for technical support during the later CUNNINGHAM, C.W. 1997. Is congruence between data partitions a stages of this project and to the authors of the “Atlas de Andalu- reliable predictor of phylogenetic accuracy? Empirically testing cia” (J. P.Gonzalez, J. M. Barnestein and D. Donaire) for all the an iterative procedure for choosing among phylogenetic methods. Systematic Biology 46, 464–478. samples of Pleurodeles from Southern Spain used in this study. DUGGEN,S.,HOERNLE,K.,VA N D E N BOGAARD,P.,RUPKE¨ ,L.& We would also like to thank Edward Wade, Soumia Fahd, R. MORGAN, J.P. 2003. Deep roots of the Messinian salinity crisis. Fonoll and David Herbert for their help and advice in various Nature 422, 602–606. matters concerning this work. We are also very grateful to J. ESTES, R. 1981. Gymnophiona, Caudata. Handbuch der Palaoherpetologie¨ Encyclopedia of Paleoherpetology Bartley and C. Griffin who run the sequencing facility at the ( ), 2 (2). Gustav Fischer Verlag, Stuttgart. Natural History Museum, London. This project was suppor- FARRIS, J.S., KALLERSJO,M.,KLUGE,A.G.&BULT, C. 1994. Testing ted by grants from the Natural Environment Research Council significance of incongruence. Cladistics 10, 315–319. (NERC) NER/B/S/2000/00714 and NER/A/S/2001/00511. FELSENSTEIN, J. 1985. Confidence-limits on phylogenies – an approach using the bootstrap. Evolution 39, 783–791. GARCIA-PARÍS,M.&JOCKUSCH, E.L. 1999. A mitochondrial DNA References perspective on the evolution of Iberian Discoglossus (Amphibia: Anura). Journal of Zoology 248, 209–218. ADAMS, C.G., GENTRY,A.W.&WHYBROW, P.S. 1983. Dating the GARCIA-PARÍS,M.,ALCOBENDAS,M.&ALBERCH, P. 1998. Influence terminal Tethyan event. In: MEULENKAMP, J. E., Ed., Reconstruc- of the Guadalquivir River basin on mitochondrial DNA evolution tion of Marine Paleoenvironments. Utrecht Micropaleontological of Salamandra salamandra (Caudata: Salamandridae) from south- Bulletin 30, 273–298. ern Spain. Copeia 173–176. ARNTZEN,J.W.&GARCIA-PARÍS, M. 1995. Morphological and al- GIULI,C.DE &TORRE, D. 1984. Species relationships and evolution lozyme studies of midwife toads (genus Alytes), including the in the Pliocene endemic faunas of Apricena (Gargano Peninsula – description of two new taxa from Spain. Contributions to Zoology Italy). Geobios Memoire´ Special (Lyon) 8, 379–383. 65, 5–34. HERRE, W. 1935. Die Schwanzlurche der mitteleocanen¨ (ober- AVISE, J.C. 2000. Phylogeography, the History and Formation of Spe- lutetischen) Braunkohle des Geiseltales und die Phylogenie cies. Cambridge, MA: Harvard University Press. der Urodelen unter Einschluss der Fossilen Formen. Zoologica BAILON, S. 1989. Les amphibiens et les reptiles du Pliocene` Superieur´ (Stuttgart) 33, 1–85. de Balaruc II (Herault,´ France). Palaeovertebrata-Montpellier 19, HERRE, W. 1941. Palaeopleurodeles hauffi nov. gen., nov. spec., ein 7–28. fossiler Schwanzlurch aus dem Miozan¨ Suddeutschlands.¨ Zoolo- BEERLI,P.,HOTZ,H.&UZZELL, T. 1996. Geologically dated sea gischer Anzeiger 134, 1–17. barriers calibrate a protein clock for Aegean water frogs. Evolution HEWITT, G.M. 1996. Some genetic consequences of ice ages, and 50, 1676–1687. their role in divergence and speciation. Biological Journal of the BLONDEL,J.&ARONSON, J. 1999. Biology and Wildlife of the Medi- Linnean Society 58, 247–276. terranean Region. Oxford University Press, New York. HEWITT, G.M. 2000. The genetic legacy of the Quaternary ice ages. BONS,J.&GENIEZ, P. 1996. Amphibians and reptiles of Morocco Nature 405, 907–913. (including Western Sahara), Biogeographical atlas. Asociacion´ HSU¨, K.J. 1972. When the Mediterranean dried up. Scientific American Herpetologica´ Espanola,˜ Barcelona. 227, 27–36. BUTLER, M. 1980. The giant erinaceid insectivore, Deinogalerix HSU¨, K.J., RYAN, W.B.F. & CITA, M.B. 1973. Late Miocene desicca- Freudental, from the Upper Miocene of Gargano, Italy. Scripta tion of the Mediterranean. Nature 242, 240–244. Geologica 57, 1–72. HSU¨, K.J., MONTADERT,L.,BEERNOUILLI,D.,CITA, M.B., ERICKSON, CACCONE,A.,MILINKOVITCH, M.C., SBORDONI,V.&POWELL,J.R. A., GARRISON, R.E., KIDD, R.B., MELIERES,F.,MULLER,C.& 1994. Molecular biogeography – using the Corsica–Sardinia mi- WRIGHT, R. 1977. History of the Mediterranean salinity crisis. croplate disjunction to calibrate mitochondrial rDNA evolutionary Nature 267, 399–403. rates in mountain newts (Euproctus). Journal of Evolutionary Bio- HUELSENBECK,J.P.&CRANDALL, K.A. 1997. Phylogeny estimation logy 7, 227–245. and hypothesis testing using maximum likelihood. Annual Review CACCONE,A.,MILINKOVITCH, M.C., SBORDONI,V.&POWELL,J.R. of Ecology and Systematics 28, 437–466. 1997. Mitochondrial DNA rates and biogeography in European HUELSENBECK,J.P.&RONQUIST, F. 2001. MRBAYES: Bayesian in- newts (Genus Euproctus). Systematic Biology 46, 126–144. ference of phylogeny. Bioinformatics 17, 754–755. CARRANZA,S.&ARNOLD, E.N. 2003. Investigating the origin of KOCHER, T.D., THOMAS, W.K., MEYER,A.,EDWARDS, S.V., PAABO, transoceanic distributions: mtDNA shows Mabuya lizards (Rep- S., VILLABLANCA,F.X.&WILSON, A.C. 1989. Dynamics of tilia, Scincidae) crossed the Atlantic twice. Systematics and Bio- mitochondrial DNA evolution in animals: amplification and se- diversity 1, 275–282. quencing with conserved primers. Proceedings of the National CARRANZA,S.,ARNOLD, E.N., THOMAS, R.H., MATEO,J.A.&LOPEZ´ - Academy of Sciences USA 86, 6196–6200. JURADO, L.F. 1999. Status of the extinct giant lacertid lizard Gallo- KRIJGSMAN,W.,GARCES,M.,AGUSTÍ,J.,RAFFI,I.,TABERNER,C. tia simonyi simonyi (Reptilia: Lacertidae) assessed using mtDNA &ZACHARIASSE, W.J. 2000. The Tortonian salinity crisis of the sequences from museum specimens. Herpetological Journal 9, eastern Betics (Spain). Earth and Planetary Science Letters 1981, 83–86. 497–511. Phylogeography of Pleurodeles 337

KRIJGSMAN,W.,HILGEN, F.J., RAFFI,I.,SIERRO,F.J.&WILSON, SAID, R. 1993. The River Nile: Geology, Hydrology and Utilization. D.S. 1999a. Chronology, causes and progression of the Messinian Oxford, Pergamon. Salinity Crisis. Nature 400, 652–655. SANCHIZ, F. 1977. Nuevos anfibios del Neogeno y Cuaternario KRIJGSMAN,W.&LANGEREIS, C.G. 2000. Magnetostratigraphy de Europa. Origen, desarrollo y relaciones de la batracofauna of the Zobzit and Koudiat Zarga sections (Taza-Guercif basin, Espanola.˜ Universidad Complutense, Madrid, tesis doctoral, Morocco): implications for the evolution of the Rifian corridor. 3 vols. 863 pp. Mar. Pet. Geol. 17, 359–371. SANDERSON, M.J. 1997. A nonparametric approach to estimating di- KRIJGSMAN,W.,LANGEREIS, C.G., ZACHARIASSE, W.J., vergence times in the absence of rate constancy. Molecular Biology BOCCALETTI,M.,MORATTI,G.,GELATI,R.,IACCARINO,S., and Evolution 14, 1218–1231. PAPANI,G.&VILLA, G. 1999b. Late Neogene evolution of the STEINFARTZ, S.U., HWANG,W.,TAUTZ,D.,O¨ Z,M.&VEITH, M. 2002. Taza-Guercif Basin (Rifian Corridor, Morocco) and implications Molecular phylogeny of the salamandrid genus Neurergus: evid- for the Messinian Salinity Crisis. Marine Geology 153, 147– ence for an intrageneric switch of reproductive biology. Amphibia– 160. Reptilia 23, 419–431. MICKEVICH,M.F.&FARRIS, J.S. 1981. The implications of congru- SWOFFORD, D.L. 1998. PAUP*: phylogenetic analysis using parsi- ence in Menidia. Systematic Zoology 30, 351–370. mony (and other methods), v 4.0 Sunderland, MA: Sinauer Asso- MILNER, A.R. 2000. Mesozoic and Tertiary Caudata and Alban- ciates. erpetontidae. In: HEATWOLE,H.&CARROL,R.L.,Eds.Am- TAN,A.M.&WAKE, D.B. 1995. MtDNA phylogeography of the phibian Biology. Surrey Beatty & Sons, Australia, pp. 1412– California newt, Taricha torosa (Caudata, Salamandridae). 1444. Molecular Phylogenetics and Evolution 14, 414–422. PASMANS,F.,BOGAERTS,S.,DONAIRE,D.&HERBERT, D. 2002. Field THOMPSON, J.D., GIBSON, T.J., PLEWNIAK,F.,JEANMOUGIN,F.& notes on Pleurodeles poireti in Tunisia. Zeitschrift fur Feldherpet- HIGGINS, D.G. 1997. The ClustalX windows interface: flexible ologie 9, 111–114. strategies for multiple sequence alignment aided by quality ana- PASTEUR, G. 1958. Sur la systematique´ des especes` du genre Pleur- lysis tools. Nucleic Acids Research 24, 4876–4882. odeles (Salamandrides).´ Bulletin de la Societ´ e´ des Sciences TITUS,T.A.&LARSON, A. 1995. A molecular phylogenetic perspect- Naturelles et Physiques du Maroc 38, 157–165. ive on the evolutionary radiation of the salamander family Sala- POSADA,D.&CRANDALL, K.A. 1998. MODELTEST: testing the mandridae. Systematic Biology 44, 125–151. model of DNA substitution. Bioinformatics 14, 817–818. VEITH,M.,KOSUCH,J.&VENCES, M. 2003. Climatic oscillations RANDO, J.C., CARRANZA,S.,MACA-MEYER,N.,ARNOLD,E.N.& triggered post-Messinian speciation of Western Palearctic brown CABRERA, V.M. In press. Status and relationships of the extinct frogs (Amphibia, Ranidae). Molecular Phylogenetics and Evolu- giant Canary Island lizard Gallotia goliath (Reptilia: Lacertidae), tion 26, 310–327. assessed using ancient mtDNA from its mummified remains. Bio- WEIJERMARS, R. 1988. Neogene tectonics in the western Mediter- logical Journal of the Linnean Society. ranean may have caused the Messinian salinity crisis and associ- ROGL¨ , F. 1999. Mediterranean and Paratethys palaeogeography dur- ated glacial events. Tectonophysics 148, 211–219. ing the Oligocene and Miocene. In AGUSTÍ,J.,ROOK,L.& WEIJERMARS, R. 1991. Geology and tectonics of the Betic zone, SE ANDREWS,P.,Eds.The Evolution of Neogene Terrestrial Spain. Earth-Science Reviews 31, 153–236. Ecosystems. Cambridge University Press, Cambridge, pp. 8– WESTPHAL, F. 1978. Tylototriton (Amphibia, Urodela) aus dem Ober- 22. miozan¨ von Ohningen. Neues Jahrbuch fur¨ Geologie und Palae- SAID, R. 1990. Geology of Egypt. Balkema, Rotterdam. ontologie Monatshefte 8, 491–501.